finding max value in a matrix

I have a matrix with 5 columns and one million rows. How can I find the max value of every 60 rows in that rows of million

Answer 1

[Jan]

Reshape the array to a 3D array with 60 elements in the first dimension. Then find the maximum. The most work is required to consider that the 1st dimension is not a multiple of 60:

data = rand(1e6, 5);
siz  = size(data);
s1   = siz(1) - mod(siz(1), 6);
x    = reshape(data(1:s1, :), 60, s1/60, siz(2));
result = max(x, [], 1);

[Matt J]

Yep, that's the thing to do. I happily accept this on AA's behalf :D

[Matt J]

In this FEX submission,

http://www.mathworks.com/matlabcentral/fileexchange/48089-separable-block-wise-operations

I generalized this to N-dimensions and to other operations with similar separability properties to max().

Answer 2

[Guillaume]

I would split the matrix into a cell array of 60 rows with mat2cell and then use cellfun (or a loop) to find the max of each cell:

 numsplit = ceil(size(m, 1) / 60); %how many cells are needed
 rowdist = ones(1, numsplit) * 60; %define the row split
 rowdist(end) = mod(size(m, 1), 60); %adjust last distance to be the modulo of row by 60
 c = mat2cell(m, rowdist, size(m, 2)); %create a cell array where each cell is 60 rows
 if isempty(c{end}) %will happen when size(m, 1) is a multiple of 60
   c(end) = [];
 end
 mx = cellfun(@(mm) max(mm(:)), c);

[Matt J]

MAT2TILES ( Download ) will take care of a lot of this book-keeping for you,

    mx = cellfun(@(mm) max(mm(:)),  mat2tiles(m,60,1));

[Guillaume]

It probably should read:

mx = cellfun(@(mm) max(mm(:)),  mat2tiles(m,60,Inf));

[Matt J]

If that's what the OP intended, then yes. You could also do

mx = cellfun(@(mm) max(mm(:)),  mat2tiles(m,60) );

[Jan]

You do not need the indirection over a cell. A 3D array is enough and much faster.